Interfacing with Computer Narratives: Literary Possibilities for Interactive Fiction
Existing interactive fiction, whether a hypertext document, a text-based system that allows for full sentence responses, or a graphical adventure with a complex interface, consists of a set of works displayed in an order determined by the interactor's input. In an Infocom interactive fiction, these are text fragments which are displayed as the interactor types commands; sometimes a series of commands are needed to elicit a certain fragment, or the interactor may need to go to one area, press a button or throw a switch, and then return to another to see what change has transpired. Perhaps the most cynical way of looking at the computer's process during an Infocom interactive fiction is as a textual shell game in which the computer responds to the interactor by shuffling pre-written blocks of text around and occasionally revealing the result. Another troubling feature of interactive fiction is the way it is centered around the solution to puzzles, giving it the "read-once" quality discussed earlier.
In response to a suggestion (originally in one of my posts) that interactive fiction's potential is greater than this puzzle-oriented model of text fragments, Graham Nelson, author of the all-text UNIX interactive fiction Curses, posted this to the USENET newsgroup rec.arts.int-fiction:
Gareth Rees [posted this] on the apparently static state of the art:
I think that there's a genuine point to be made, which is that text-only Interactive Fiction is in a bit of a creative doldrums at the moment, i.e. there isn't much in the way of creative or artistic innovation. ... it seems that IF authors have a model for IF that we know works (i.e. a prologue/middle game/endgame structure, a hierarchical object-based world model, a strongly goal-oriented plot whose solution is repeatedly blocked by puzzles), but no-one really knows what to do next.
... I feel that although the classical model is indeed long-established, it's important to learn it inside out before going on to more advanced things: the same way that poets ought to learn to write sonnets before attempting free verse.
This evinces some interest in expanding the current boundaries of interactive fiction, but it reveals that interactive fiction designers still focus on puzzle-based narrative. At least this focus is being questioned. In the newsgroup for interactive fiction designers in which the above message appeared, there is no questioning of the basic assumption that interactive fiction must be composed of pre-constructed fragments of text or other media. The textual "shell game" is still the unchallenged framework for interactive fiction.
Just as computers can parse sentences typed by people to discern the semantics (with some degree of success), computers can also generate sentences. Interactive fiction is not constrained to be a series of sentence-long fragments displayed according to rules. Rather, text produced by the computer, word by word, according to the designer's rules, could constitute the contribution of the software to an interactive narrative.
Calvino's literature machine, which produces all works by enumerating all strings of a particular language, is not fantasy or speculation. Given a particular length in characters, it is trivial for a programmer to set a computer about the task of enumerating all strings of characters of that length. The prohibitive amount of time required is the sole barrier to the automatic generation of Jorge Luis Borges' Library of Babylon, in which every possible book of a particular length existed. Technologically, these texts could be enumerated. Among the works, if the length of the strings was initially set correctly, would be Hamlet, produced as if by a roomful of monkeys with typewriters.
Of course, the usefulness of such enumeration is destroyed because of the problem of culling literature, or even sensible language, from billions of garbled strands of characters. It is useless to have a computer produce strings of characters without being able to discern which of those are valid strings of English. If the computer starts with words, rather than characters, as atomic units, it is likelier to arrive at a valid sentence, but the probability is still minute. Rather than producing strings of words and checking to see if these are meaningful, using encoded rules of grammar has seen more success as a technique for producing language. Although it is a simplification to see these rules as a reversal of the way that a program (like the Interlogic parser) dissects language in order to extract semantics, there are similarities.
Much work in creative text generation has been done. This work, some of which has been very successful, ranges from the efforts of a computer hobbyist to write a program that will produce romance novels to computer scientists' efforts to generate serial literature ("'soap opera' type stories") or generate prose for interactive fiction in such as way that the user can adjust traits like mood, voice, and formality with software "knobs." (Lohr; Lebowitz 171; Kantrowitz 1)
The partial success of story-generating software is heartening, but interactive fiction that generates text need not be founded on these successes. Romance novels and 'soap opera'-like serial fiction are products; interactive fiction is a process. Just as a transcript of an enjoyable conversation may not be good reading, the texts produced during an interaction with an interactive fiction do not have to be publishable or serve as television scripts. Because the interactor's process is as important as the computer's, the burden on the computer as a generator of language is reduced.
Of course, a program that merely fills in blanks with random words is not a part of a very literary process. As designers craft rules for higher-level structures in language and make the language generated responsive to the input of the interactor, however, the potential for literariness is enhanced. When interactive fiction becomes a generator of text rather than just a method of displaying text, the ultimate possibility of the medium in unlocked. "[I]nteractive fiction is, in principle (if not in practice), open-ended - infinite." By generating a never-ending story, the problem of death and other endings is solved. The fiction and the interactor can "create a story within a story or a story that generates another story within itself which generates another story within itself and so on, fictions dizzying and dazzling." This "hall-of-mirrors" work would be sui generis. (Niesz and Holland 121)